Identifying and extracting values of data fields from images of documents requires data and resource intensive optical character recognition processes. The processing power and time required to identify and extract important information from images of documents is significantly reduced if the OCR processes are tailored to specific data fields of the image of the documents and applied only to regions of the document images where that data field is located. However, as the volume and variety of documents increases, the complexity of the OCR process systems increases significantly. Additionally, as some resource documents include data fields in positions that are outliers to default or expected coordinate areas on the document image, manual entry of the values of these data fields is time consuming and exposes the processing system to user error.
Therefore, a need exists to dynamically tune the OCR process for identifying and extracting values of data fields from images of resource documents. The embodiments of the invention herein are designed to quickly correct the image coordinate regions for outlier data fields and store these updated coordinate regions so they can be used to automatically process subsequently received images of documents from the same source.